Decremental User Selection for Large-Scale Multi-User MIMO Downlink with Zero-Forcing Beamforming

This paper proposes a decremental user selection algorithm based on zero-forcing beamforming when the number of users K in the network is smaller than the number of antennas M at a base station. The algorithm is specifically designed for large-scale multi-user multiple-input multiple-output (MIMO) downlink channels. While previous user selection algorithms are based on incremental search that starts from an empty user set, our proposed delete the minimum lambda (DML) algorithm starts by selecting all users and then deleting one user per iteration. DML substantially reduces the computational complexity as the cardinality of the final user set is close to K. Simulation results indicate that on average DML achieves an equal or higher sum rate performance than previous algorithms with greatly reduced complexity of O(MK²).